/* SPDX-License-Identifier: BSD-3-Clause
- * Copyright(c) 2018-2019 Hisilicon Limited.
+ * Copyright(c) 2018-2021 HiSilicon Limited.
*/
#ifndef _HNS3_RXTX_H_
#define _HNS3_RXTX_H_
+#include <stdint.h>
+#include <rte_mbuf_core.h>
+
#define HNS3_MIN_RING_DESC 64
#define HNS3_MAX_RING_DESC 32768
#define HNS3_DEFAULT_RING_DESC 1024
#define HNS3_DEFAULT_TX_RS_THRESH 32
#define HNS3_TX_FAST_FREE_AHEAD 64
-#define HNS3_DEFAULT_RX_BURST 32
+#define HNS3_DEFAULT_RX_BURST 64
#if (HNS3_DEFAULT_RX_BURST > 64)
#error "PMD HNS3: HNS3_DEFAULT_RX_BURST must <= 64\n"
#endif
#define HNS3_RXD_OL3ID_M (0xf << HNS3_RXD_OL3ID_S)
#define HNS3_RXD_OL4ID_S 8
#define HNS3_RXD_OL4ID_M (0xf << HNS3_RXD_OL4ID_S)
+#define HNS3_RXD_PTYPE_S 4
+#define HNS3_RXD_PTYPE_M (0xff << HNS3_RXD_PTYPE_S)
#define HNS3_RXD_FBHI_S 12
#define HNS3_RXD_FBHI_M (0x3 << HNS3_RXD_FBHI_S)
#define HNS3_RXD_FBLI_S 14
#define HNS3_RXD_LUM_B 9
#define HNS3_RXD_CRCP_B 10
#define HNS3_RXD_L3L4P_B 11
-#define HNS3_RXD_TSIND_S 12
-#define HNS3_RXD_TSIND_M (0x7 << HNS3_RXD_TSIND_S)
-#define HNS3_RXD_LKBK_B 15
+
+#define HNS3_RXD_TS_VLD_B 14
#define HNS3_RXD_GRO_SIZE_S 16
#define HNS3_RXD_GRO_SIZE_M (0x3fff << HNS3_RXD_GRO_SIZE_S)
#define HNS3_TXD_MSS_S 0
#define HNS3_TXD_MSS_M (0x3fff << HNS3_TXD_MSS_S)
+#define HNS3_TXD_OL4CS_B 22
#define HNS3_L2_LEN_UNIT 1UL
#define HNS3_L3_LEN_UNIT 2UL
#define HNS3_L4_LEN_UNIT 2UL
struct hns3_desc {
union {
uint64_t addr;
+ uint64_t timestamp;
+
struct {
uint32_t addr0;
uint32_t addr1;
};
};
- uint32_t paylen;
+ uint32_t paylen_fd_dop_ol4cs;
uint16_t tp_fe_sc_vld_ra_ri;
uint16_t mss;
} tx;
struct rte_mbuf *mbuf;
};
+struct hns3_rx_basic_stats {
+ uint64_t packets;
+ uint64_t bytes;
+ uint64_t errors;
+};
+
+struct hns3_rx_dfx_stats {
+ uint64_t l3_csum_errors;
+ uint64_t l4_csum_errors;
+ uint64_t ol3_csum_errors;
+ uint64_t ol4_csum_errors;
+};
+
+struct hns3_rx_bd_errors_stats {
+ uint64_t l2_errors;
+ uint64_t pkt_len_errors;
+};
+
struct hns3_rx_queue {
- void *io_base;
volatile void *io_head_reg;
- struct hns3_adapter *hns;
struct hns3_ptype_table *ptype_tbl;
struct rte_mempool *mb_pool;
struct hns3_desc *rx_ring;
- uint64_t rx_ring_phys_addr; /* RX ring DMA address */
- const struct rte_memzone *mz;
struct hns3_entry *sw_ring;
- struct rte_mbuf *pkt_first_seg;
- struct rte_mbuf *pkt_last_seg;
- uint16_t queue_id;
uint16_t port_id;
uint16_t nb_rx_desc;
- uint16_t rx_buf_len;
/*
* threshold for the number of BDs waited to passed to hardware. If the
* number exceeds the threshold, driver will pass these BDs to hardware.
uint16_t rx_rearm_start; /* index of BD that driver re-arming from */
uint16_t rx_rearm_nb; /* number of remaining BDs to be re-armed */
- /* 4 if DEV_RX_OFFLOAD_KEEP_CRC offload set, 0 otherwise */
+ /* 4 if RTE_ETH_RX_OFFLOAD_KEEP_CRC offload set, 0 otherwise */
uint8_t crc_len;
- bool rx_deferred_start; /* don't start this queue in dev start */
- bool configured; /* indicate if rx queue has been configured */
/*
* Indicate whether ignore the outer VLAN field in the Rx BD reported
* by the Hardware. Because the outer VLAN is the PVID if the PVID is
* should not be transitted to the upper-layer application. For hardware
* network engine whose vlan mode is HNS3_HW_SHIFT_AND_DISCARD_MODE,
* such as kunpeng 930, PVID will not be reported to the BDs. So, PMD
- * driver does not need to perform PVID-related operation in Rx. At this
+ * does not need to perform PVID-related operation in Rx. At this
* point, the pvid_sw_discard_en will be false.
*/
- bool pvid_sw_discard_en;
+ uint8_t pvid_sw_discard_en:1;
+ uint8_t ptype_en:1; /* indicate if the ptype field enabled */
- uint64_t l2_errors;
- uint64_t pkt_len_errors;
- uint64_t l3_csum_errors;
- uint64_t l4_csum_errors;
- uint64_t ol3_csum_errors;
- uint64_t ol4_csum_errors;
+ uint64_t mbuf_initializer; /* value to init mbufs used with vector rx */
+ /* offset_table: used for vector, to solve execute re-order problem */
+ uint8_t offset_table[HNS3_VECTOR_RX_OFFSET_TABLE_LEN + 1];
+
+ uint16_t bulk_mbuf_num; /* indicate bulk_mbuf valid nums */
+
+ struct hns3_rx_basic_stats basic_stats;
+
+ struct rte_mbuf *pkt_first_seg;
+ struct rte_mbuf *pkt_last_seg;
struct rte_mbuf *bulk_mbuf[HNS3_BULK_ALLOC_MBUF_NUM];
- uint16_t bulk_mbuf_num;
- /* offset_table: used for vector, to solve execute re-order problem */
- uint8_t offset_table[HNS3_VECTOR_RX_OFFSET_TABLE_LEN + 1];
- uint64_t mbuf_initializer; /* value to init mbufs used with vector rx */
+ /* DFX statistics that driver does not need to discard packets */
+ struct hns3_rx_dfx_stats dfx_stats;
+ /* Error statistics that driver needs to discard packets */
+ struct hns3_rx_bd_errors_stats err_stats;
+
struct rte_mbuf fake_mbuf; /* fake mbuf used with vector rx */
+
+ /*
+ * The following fields are not accessed in the I/O path, so they are
+ * placed at the end.
+ */
+ void *io_base;
+ struct hns3_adapter *hns;
+ uint64_t rx_ring_phys_addr; /* RX ring DMA address */
+ const struct rte_memzone *mz;
+
+ uint16_t queue_id;
+ uint16_t rx_buf_len;
+
+ bool configured; /* indicate if rx queue has been configured */
+ bool rx_deferred_start; /* don't start this queue in dev start */
+ bool enabled; /* indicate if Rx queue has been enabled */
+};
+
+struct hns3_tx_basic_stats {
+ uint64_t packets;
+ uint64_t bytes;
+};
+
+/*
+ * The following items are used for the abnormal errors statistics in
+ * the Tx datapath. When upper level application calls the
+ * rte_eth_tx_burst API function to send multiple packets at a time with
+ * burst mode based on hns3 network engine, there are some abnormal
+ * conditions that cause the driver to fail to operate the hardware to
+ * send packets correctly.
+ * Note: When using burst mode to call the rte_eth_tx_burst API function
+ * to send multiple packets at a time. When the first abnormal error is
+ * detected, add one to the relevant error statistics item, and then
+ * exit the loop of sending multiple packets of the function. That is to
+ * say, even if there are multiple packets in which abnormal errors may
+ * be detected in the burst, the relevant error statistics in the driver
+ * will only be increased by one.
+ * The detail description of the Tx abnormal errors statistic items as
+ * below:
+ * - over_length_pkt_cnt
+ * Total number of greater than HNS3_MAX_FRAME_LEN the driver
+ * supported.
+ *
+ * - exceed_limit_bd_pkt_cnt
+ * Total number of exceeding the hardware limited bd which process
+ * a packet needed bd numbers.
+ *
+ * - exceed_limit_bd_reassem_fail
+ * Total number of exceeding the hardware limited bd fail which
+ * process a packet needed bd numbers and reassemble fail.
+ *
+ * - unsupported_tunnel_pkt_cnt
+ * Total number of unsupported tunnel packet. The unsupported tunnel
+ * type: vxlan_gpe, gtp, ipip and MPLSINUDP, MPLSINUDP is a packet
+ * with MPLS-in-UDP RFC 7510 header.
+ *
+ * - queue_full_cnt
+ * Total count which the available bd numbers in current bd queue is
+ * less than the bd numbers with the pkt process needed.
+ *
+ * - pkt_padding_fail_cnt
+ * Total count which the packet length is less than minimum packet
+ * length(struct hns3_tx_queue::min_tx_pkt_len) supported by
+ * hardware in Tx direction and fail to be appended with 0.
+ */
+struct hns3_tx_dfx_stats {
+ uint64_t over_length_pkt_cnt;
+ uint64_t exceed_limit_bd_pkt_cnt;
+ uint64_t exceed_limit_bd_reassem_fail;
+ uint64_t unsupported_tunnel_pkt_cnt;
+ uint64_t queue_full_cnt;
+ uint64_t pkt_padding_fail_cnt;
};
struct hns3_tx_queue {
- void *io_base;
+ /* The io_tail_reg is write-only if working in tx push mode */
volatile void *io_tail_reg;
- struct hns3_adapter *hns;
struct hns3_desc *tx_ring;
- uint64_t tx_ring_phys_addr; /* TX ring DMA address */
- const struct rte_memzone *mz;
struct hns3_entry *sw_ring;
- uint16_t queue_id;
- uint16_t port_id;
uint16_t nb_tx_desc;
/*
* index of next BD whose corresponding rte_mbuf can be released by
uint16_t tx_free_thresh;
/*
- * For better performance in tx datapath, releasing mbuf in batches is
- * required.
- * Only checking the VLD bit of the last descriptor in a batch of the
- * thresh descriptors does not mean that these descriptors are all sent
- * by hardware successfully. So we need to check that the VLD bits of
- * all descriptors are cleared. and then free all mbufs in the batch.
- * - tx_rs_thresh
- * Number of mbufs released at a time.
- *
- * - free
- * Tx mbuf free array used for preserving temporarily address of mbuf
- * released back to mempool, when releasing mbuf in batches.
+ * The minimum length of the packet supported by hardware in the Tx
+ * direction.
*/
- uint16_t tx_rs_thresh;
- struct rte_mbuf **free;
+ uint8_t min_tx_pkt_len;
+
+ uint8_t max_non_tso_bd_num; /* max BD number of one non-TSO packet */
/*
- * The minimum length of the packet supported by hardware in the Tx
- * direction.
+ * tso mode.
+ * value range:
+ * HNS3_TSO_SW_CAL_PSEUDO_H_CSUM/HNS3_TSO_HW_CAL_PSEUDO_H_CSUM
+ *
+ * - HNS3_TSO_SW_CAL_PSEUDO_H_CSUM
+ * In this mode, because of the hardware constraint, network driver
+ * software need erase the L4 len value of the TCP pseudo header
+ * and recalculate the TCP pseudo header checksum of packets that
+ * need TSO.
+ *
+ * - HNS3_TSO_HW_CAL_PSEUDO_H_CSUM
+ * In this mode, hardware support recalculate the TCP pseudo header
+ * checksum of packets that need TSO, so network driver software
+ * not need to recalculate it.
+ */
+ uint16_t tso_mode:1;
+ /*
+ * udp checksum mode.
+ * value range:
+ * HNS3_SPECIAL_PORT_HW_CKSUM_MODE/HNS3_SPECIAL_PORT_SW_CKSUM_MODE
+ *
+ * - HNS3_SPECIAL_PORT_SW_CKSUM_MODE
+ * In this mode, HW can not do checksum for special UDP port like
+ * 4789, 4790, 6081 for non-tunnel UDP packets and UDP tunnel
+ * packets without the RTE_MBUF_F_TX_TUNEL_MASK in the mbuf. So, PMD need
+ * do the checksum for these packets to avoid a checksum error.
+ *
+ * - HNS3_SPECIAL_PORT_HW_CKSUM_MODE
+ * In this mode, HW does not have the preceding problems and can
+ * directly calculate the checksum of these UDP packets.
*/
- uint32_t min_tx_pkt_len;
+ uint16_t udp_cksum_mode:1;
- bool tx_deferred_start; /* don't start this queue in dev start */
- bool configured; /* indicate if tx queue has been configured */
+ uint16_t simple_bd_enable:1;
+ uint16_t tx_push_enable:1; /* check whether the tx push is enabled */
/*
* Indicate whether add the vlan_tci of the mbuf to the inner VLAN field
* of Tx BD. Because the outer VLAN will always be the PVID when the
* PVID will overwrite the outer VLAN field of Tx BD. For the hardware
* network engine whose vlan mode is HNS3_HW_SHIFT_AND_DISCARD_MODE,
* such as kunpeng 930, if the PVID is set, the hardware will shift the
- * VLAN field automatically. So, PMD driver does not need to do
+ * VLAN field automatically. So, PMD does not need to do
* PVID-related operations in Tx. And pvid_sw_shift_en will be false at
* this point.
*/
- bool pvid_sw_shift_en;
+ uint16_t pvid_sw_shift_en:1;
+ /* check whether the mbuf fast free offload is enabled */
+ uint16_t mbuf_fast_free_en:1;
/*
- * The following items are used for the abnormal errors statistics in
- * the Tx datapath. When upper level application calls the
- * rte_eth_tx_burst API function to send multiple packets at a time with
- * burst mode based on hns3 network engine, there are some abnormal
- * conditions that cause the driver to fail to operate the hardware to
- * send packets correctly.
- * Note: When using burst mode to call the rte_eth_tx_burst API function
- * to send multiple packets at a time. When the first abnormal error is
- * detected, add one to the relevant error statistics item, and then
- * exit the loop of sending multiple packets of the function. That is to
- * say, even if there are multiple packets in which abnormal errors may
- * be detected in the burst, the relevant error statistics in the driver
- * will only be increased by one.
- * The detail description of the Tx abnormal errors statistic items as
- * below:
- * - over_length_pkt_cnt
- * Total number of greater than HNS3_MAX_FRAME_LEN the driver
- * supported.
- *
- * - exceed_limit_bd_pkt_cnt
- * Total number of exceeding the hardware limited bd which process
- * a packet needed bd numbers.
- *
- * - exceed_limit_bd_reassem_fail
- * Total number of exceeding the hardware limited bd fail which
- * process a packet needed bd numbers and reassemble fail.
- *
- * - unsupported_tunnel_pkt_cnt
- * Total number of unsupported tunnel packet. The unsupported tunnel
- * type: vxlan_gpe, gtp, ipip and MPLSINUDP, MPLSINUDP is a packet
- * with MPLS-in-UDP RFC 7510 header.
- *
- * - queue_full_cnt
- * Total count which the available bd numbers in current bd queue is
- * less than the bd numbers with the pkt process needed.
+ * For better performance in tx datapath, releasing mbuf in batches is
+ * required.
+ * Only checking the VLD bit of the last descriptor in a batch of the
+ * thresh descriptors does not mean that these descriptors are all sent
+ * by hardware successfully. So we need to check that the VLD bits of
+ * all descriptors are cleared. and then free all mbufs in the batch.
+ * - tx_rs_thresh
+ * Number of mbufs released at a time.
*
- * - pkt_padding_fail_cnt
- * Total count which the packet length is less than minimum packet
- * length(struct hns3_tx_queue::min_tx_pkt_len) supported by
- * hardware in Tx direction and fail to be appended with 0.
+ * - free
+ * Tx mbuf free array used for preserving temporarily address of mbuf
+ * released back to mempool, when releasing mbuf in batches.
*/
- uint64_t over_length_pkt_cnt;
- uint64_t exceed_limit_bd_pkt_cnt;
- uint64_t exceed_limit_bd_reassem_fail;
- uint64_t unsupported_tunnel_pkt_cnt;
- uint64_t queue_full_cnt;
- uint64_t pkt_padding_fail_cnt;
+ uint16_t tx_rs_thresh;
+ struct rte_mbuf **free;
+
+ struct hns3_tx_basic_stats basic_stats;
+ struct hns3_tx_dfx_stats dfx_stats;
+
+ /*
+ * The following fields are not accessed in the I/O path, so they are
+ * placed at the end.
+ */
+ void *io_base;
+ struct hns3_adapter *hns;
+ uint64_t tx_ring_phys_addr; /* TX ring DMA address */
+ const struct rte_memzone *mz;
+
+ uint16_t port_id;
+ uint16_t queue_id;
+
+ bool configured; /* indicate if tx queue has been configured */
+ bool tx_deferred_start; /* don't start this queue in dev start */
+ bool enabled; /* indicate if Tx queue has been enabled */
};
#define HNS3_GET_TX_QUEUE_PEND_BD_NUM(txq) \
unsigned int socket_id;
};
-#define HNS3_TX_CKSUM_OFFLOAD_MASK ( \
- PKT_TX_OUTER_IPV6 | \
- PKT_TX_OUTER_IPV4 | \
- PKT_TX_OUTER_IP_CKSUM | \
- PKT_TX_IPV6 | \
- PKT_TX_IPV4 | \
- PKT_TX_IP_CKSUM | \
- PKT_TX_L4_MASK | \
- PKT_TX_TUNNEL_MASK)
+#define HNS3_TX_CKSUM_OFFLOAD_MASK (RTE_MBUF_F_TX_OUTER_UDP_CKSUM | \
+ RTE_MBUF_F_TX_OUTER_IP_CKSUM | \
+ RTE_MBUF_F_TX_IP_CKSUM | \
+ RTE_MBUF_F_TX_TCP_SEG | \
+ RTE_MBUF_F_TX_L4_MASK)
enum hns3_cksum_status {
HNS3_CKSUM_NONE = 0,
HNS3_OUTER_L4_CKSUM_ERR = 8
};
-static inline int
-hns3_handle_bdinfo(struct hns3_rx_queue *rxq, struct rte_mbuf *rxm,
- uint32_t bd_base_info, uint32_t l234_info,
- uint32_t *cksum_err)
+extern uint64_t hns3_timestamp_rx_dynflag;
+extern int hns3_timestamp_dynfield_offset;
+
+static inline void
+hns3_rx_set_cksum_flag(struct hns3_rx_queue *rxq,
+ struct rte_mbuf *rxm,
+ uint32_t l234_info)
{
-#define L2E_TRUNC_ERR_FLAG (BIT(HNS3_RXD_L2E_B) | \
- BIT(HNS3_RXD_TRUNCATE_B))
-#define CHECKSUM_ERR_FLAG (BIT(HNS3_RXD_L3E_B) | \
+#define HNS3_RXD_CKSUM_ERR_MASK (BIT(HNS3_RXD_L3E_B) | \
BIT(HNS3_RXD_L4E_B) | \
BIT(HNS3_RXD_OL3E_B) | \
BIT(HNS3_RXD_OL4E_B))
- uint32_t tmp = 0;
+ if (likely((l234_info & HNS3_RXD_CKSUM_ERR_MASK) == 0)) {
+ rxm->ol_flags |= (RTE_MBUF_F_RX_IP_CKSUM_GOOD | RTE_MBUF_F_RX_L4_CKSUM_GOOD);
+ return;
+ }
+
+ if (unlikely(l234_info & BIT(HNS3_RXD_L3E_B))) {
+ rxm->ol_flags |= RTE_MBUF_F_RX_IP_CKSUM_BAD;
+ rxq->dfx_stats.l3_csum_errors++;
+ } else {
+ rxm->ol_flags |= RTE_MBUF_F_RX_IP_CKSUM_GOOD;
+ }
+
+ if (unlikely(l234_info & BIT(HNS3_RXD_L4E_B))) {
+ rxm->ol_flags |= RTE_MBUF_F_RX_L4_CKSUM_BAD;
+ rxq->dfx_stats.l4_csum_errors++;
+ } else {
+ rxm->ol_flags |= RTE_MBUF_F_RX_L4_CKSUM_GOOD;
+ }
+
+ if (unlikely(l234_info & BIT(HNS3_RXD_OL3E_B)))
+ rxq->dfx_stats.ol3_csum_errors++;
+
+ if (unlikely(l234_info & BIT(HNS3_RXD_OL4E_B))) {
+ rxm->ol_flags |= RTE_MBUF_F_RX_OUTER_L4_CKSUM_BAD;
+ rxq->dfx_stats.ol4_csum_errors++;
+ }
+}
+
+static inline int
+hns3_handle_bdinfo(struct hns3_rx_queue *rxq, struct rte_mbuf *rxm,
+ uint32_t bd_base_info, uint32_t l234_info)
+{
+#define L2E_TRUNC_ERR_FLAG (BIT(HNS3_RXD_L2E_B) | \
+ BIT(HNS3_RXD_TRUNCATE_B))
/*
* If packet len bigger than mtu when recv with no-scattered algorithm,
- * the first n bd will without FE bit, we need process this sisution.
+ * the first n bd will without FE bit, we need process this situation.
* Note: we don't need add statistic counter because latest BD which
* with FE bit will mark HNS3_RXD_L2E_B bit.
*/
if (unlikely((l234_info & L2E_TRUNC_ERR_FLAG) || rxm->pkt_len == 0)) {
if (l234_info & BIT(HNS3_RXD_L2E_B))
- rxq->l2_errors++;
+ rxq->err_stats.l2_errors++;
else
- rxq->pkt_len_errors++;
+ rxq->err_stats.pkt_len_errors++;
return -EINVAL;
}
- if (bd_base_info & BIT(HNS3_RXD_L3L4P_B)) {
- if (likely((l234_info & CHECKSUM_ERR_FLAG) == 0)) {
- *cksum_err = 0;
- return 0;
- }
-
- if (unlikely(l234_info & BIT(HNS3_RXD_L3E_B))) {
- rxm->ol_flags |= PKT_RX_IP_CKSUM_BAD;
- rxq->l3_csum_errors++;
- tmp |= HNS3_L3_CKSUM_ERR;
- }
-
- if (unlikely(l234_info & BIT(HNS3_RXD_L4E_B))) {
- rxm->ol_flags |= PKT_RX_L4_CKSUM_BAD;
- rxq->l4_csum_errors++;
- tmp |= HNS3_L4_CKSUM_ERR;
- }
-
- if (unlikely(l234_info & BIT(HNS3_RXD_OL3E_B))) {
- rxq->ol3_csum_errors++;
- tmp |= HNS3_OUTER_L3_CKSUM_ERR;
- }
-
- if (unlikely(l234_info & BIT(HNS3_RXD_OL4E_B))) {
- rxm->ol_flags |= PKT_RX_OUTER_L4_CKSUM_BAD;
- rxq->ol4_csum_errors++;
- tmp |= HNS3_OUTER_L4_CKSUM_ERR;
- }
- }
- *cksum_err = tmp;
+ if (bd_base_info & BIT(HNS3_RXD_L3L4P_B))
+ hns3_rx_set_cksum_flag(rxq, rxm, l234_info);
return 0;
}
-static inline void
-hns3_rx_set_cksum_flag(struct rte_mbuf *rxm, const uint64_t packet_type,
- const uint32_t cksum_err)
-{
- if (unlikely((packet_type & RTE_PTYPE_TUNNEL_MASK))) {
- if (likely(packet_type & RTE_PTYPE_INNER_L3_MASK) &&
- (cksum_err & HNS3_L3_CKSUM_ERR) == 0)
- rxm->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
- if (likely(packet_type & RTE_PTYPE_INNER_L4_MASK) &&
- (cksum_err & HNS3_L4_CKSUM_ERR) == 0)
- rxm->ol_flags |= PKT_RX_L4_CKSUM_GOOD;
- if (likely(packet_type & RTE_PTYPE_L4_MASK) &&
- (cksum_err & HNS3_OUTER_L4_CKSUM_ERR) == 0)
- rxm->ol_flags |= PKT_RX_OUTER_L4_CKSUM_GOOD;
- } else {
- if (likely(packet_type & RTE_PTYPE_L3_MASK) &&
- (cksum_err & HNS3_L3_CKSUM_ERR) == 0)
- rxm->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
- if (likely(packet_type & RTE_PTYPE_L4_MASK) &&
- (cksum_err & HNS3_L4_CKSUM_ERR) == 0)
- rxm->ol_flags |= PKT_RX_L4_CKSUM_GOOD;
- }
-}
-
static inline uint32_t
hns3_rx_calc_ptype(struct hns3_rx_queue *rxq, const uint32_t l234_info,
const uint32_t ol_info)
{
- const struct hns3_ptype_table *const ptype_tbl = rxq->ptype_tbl;
- uint32_t l2id, l3id, l4id;
+ const struct hns3_ptype_table * const ptype_tbl = rxq->ptype_tbl;
uint32_t ol3id, ol4id;
+ uint32_t l3id, l4id;
+ uint32_t ptype;
+
+ if (rxq->ptype_en) {
+ ptype = hns3_get_field(ol_info, HNS3_RXD_PTYPE_M,
+ HNS3_RXD_PTYPE_S);
+ return ptype_tbl->ptype[ptype];
+ }
ol4id = hns3_get_field(ol_info, HNS3_RXD_OL4ID_M, HNS3_RXD_OL4ID_S);
ol3id = hns3_get_field(ol_info, HNS3_RXD_OL3ID_M, HNS3_RXD_OL3ID_S);
- l2id = hns3_get_field(l234_info, HNS3_RXD_STRP_TAGP_M,
- HNS3_RXD_STRP_TAGP_S);
l3id = hns3_get_field(l234_info, HNS3_RXD_L3ID_M, HNS3_RXD_L3ID_S);
l4id = hns3_get_field(l234_info, HNS3_RXD_L4ID_M, HNS3_RXD_L4ID_S);
if (unlikely(ptype_tbl->ol4table[ol4id]))
- return ptype_tbl->inner_l2table[l2id] |
- ptype_tbl->inner_l3table[l3id] |
+ return ptype_tbl->inner_l3table[l3id] |
ptype_tbl->inner_l4table[l4id] |
- ptype_tbl->ol3table[ol3id] | ptype_tbl->ol4table[ol4id];
+ ptype_tbl->ol3table[ol3id] |
+ ptype_tbl->ol4table[ol4id];
else
- return ptype_tbl->l2table[l2id] | ptype_tbl->l3table[l3id] |
- ptype_tbl->l4table[l4id];
+ return ptype_tbl->l3table[l3id] | ptype_tbl->l4table[l4id];
}
-void hns3_dev_rx_queue_release(void *queue);
-void hns3_dev_tx_queue_release(void *queue);
+/*
+ * If enable using Tx push feature and also device support it, then use quick
+ * doorbell (bar45) to inform the hardware.
+ *
+ * The other cases (such as: device don't support or user don't enable using)
+ * then use normal doorbell (bar23) to inform the hardware.
+ */
+static inline void
+hns3_write_txq_tail_reg(struct hns3_tx_queue *txq, uint32_t value)
+{
+ rte_io_wmb();
+ if (txq->tx_push_enable)
+ rte_write64_relaxed(rte_cpu_to_le_32(value), txq->io_tail_reg);
+ else
+ rte_write32_relaxed(rte_cpu_to_le_32(value), txq->io_tail_reg);
+}
+
+void hns3_dev_rx_queue_release(struct rte_eth_dev *dev, uint16_t queue_id);
+void hns3_dev_tx_queue_release(struct rte_eth_dev *dev, uint16_t queue_id);
void hns3_free_all_queues(struct rte_eth_dev *dev);
-int hns3_reset_all_queues(struct hns3_adapter *hns);
+int hns3_reset_all_tqps(struct hns3_adapter *hns);
void hns3_dev_all_rx_queue_intr_enable(struct hns3_hw *hw, bool en);
int hns3_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
int hns3_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
void hns3_enable_all_queues(struct hns3_hw *hw, bool en);
-int hns3_start_queues(struct hns3_adapter *hns, bool reset_queue);
-int hns3_stop_queues(struct hns3_adapter *hns, bool reset_queue);
+int hns3_init_queues(struct hns3_adapter *hns, bool reset_queue);
+void hns3_start_tqps(struct hns3_hw *hw);
+void hns3_stop_tqps(struct hns3_hw *hw);
int hns3_rxq_iterate(struct rte_eth_dev *dev,
int (*callback)(struct hns3_rx_queue *, void *), void *arg);
void hns3_dev_release_mbufs(struct hns3_adapter *hns);
struct rte_mempool *mp);
int hns3_tx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t nb_desc,
unsigned int socket, const struct rte_eth_txconf *conf);
-uint16_t hns3_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
- uint16_t nb_pkts);
+uint32_t hns3_rx_queue_count(void *rx_queue);
+int hns3_dev_rx_queue_start(struct rte_eth_dev *dev, uint16_t rx_queue_id);
+int hns3_dev_rx_queue_stop(struct rte_eth_dev *dev, uint16_t rx_queue_id);
+int hns3_dev_tx_queue_start(struct rte_eth_dev *dev, uint16_t tx_queue_id);
+int hns3_dev_tx_queue_stop(struct rte_eth_dev *dev, uint16_t tx_queue_id);
+uint16_t hns3_recv_pkts_simple(void *rx_queue, struct rte_mbuf **rx_pkts,
+ uint16_t nb_pkts);
uint16_t hns3_recv_scattered_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
uint16_t nb_pkts);
uint16_t hns3_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
uint16_t nb_pkts);
+uint16_t hns3_recv_pkts_vec_sve(void *rx_queue, struct rte_mbuf **rx_pkts,
+ uint16_t nb_pkts);
int hns3_rx_burst_mode_get(struct rte_eth_dev *dev,
__rte_unused uint16_t queue_id,
struct rte_eth_burst_mode *mode);
-int hns3_rx_check_vec_support(struct rte_eth_dev *dev);
uint16_t hns3_prep_pkts(__rte_unused void *tx_queue, struct rte_mbuf **tx_pkts,
uint16_t nb_pkts);
uint16_t hns3_xmit_pkts_simple(void *tx_queue, struct rte_mbuf **tx_pkts,
uint16_t nb_pkts);
uint16_t hns3_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
uint16_t nb_pkts);
+uint16_t hns3_xmit_pkts_vec_sve(void *tx_queue, struct rte_mbuf **tx_pkts,
+ uint16_t nb_pkts);
int hns3_tx_burst_mode_get(struct rte_eth_dev *dev,
__rte_unused uint16_t queue_id,
struct rte_eth_burst_mode *mode);
const uint32_t *hns3_dev_supported_ptypes_get(struct rte_eth_dev *dev);
void hns3_init_rx_ptype_tble(struct rte_eth_dev *dev);
void hns3_set_rxtx_function(struct rte_eth_dev *eth_dev);
+eth_tx_burst_t hns3_get_tx_function(struct rte_eth_dev *dev,
+ eth_tx_prep_t *prep);
+uint16_t hns3_dummy_rxtx_burst(void *dpdk_txq __rte_unused,
+ struct rte_mbuf **pkts __rte_unused,
+ uint16_t pkts_n __rte_unused);
+
+uint32_t hns3_get_tqp_intr_reg_offset(uint16_t tqp_intr_id);
void hns3_set_queue_intr_gl(struct hns3_hw *hw, uint16_t queue_id,
uint8_t gl_idx, uint16_t gl_value);
void hns3_set_queue_intr_rl(struct hns3_hw *hw, uint16_t queue_id,
struct rte_eth_rxq_info *qinfo);
void hns3_txq_info_get(struct rte_eth_dev *dev, uint16_t queue_id,
struct rte_eth_txq_info *qinfo);
+uint32_t hns3_get_tqp_reg_offset(uint16_t idx);
+int hns3_start_all_txqs(struct rte_eth_dev *dev);
+int hns3_start_all_rxqs(struct rte_eth_dev *dev);
+void hns3_stop_all_txqs(struct rte_eth_dev *dev);
+void hns3_restore_tqp_enable_state(struct hns3_hw *hw);
+int hns3_tx_done_cleanup(void *txq, uint32_t free_cnt);
+void hns3_enable_rxd_adv_layout(struct hns3_hw *hw);
+int hns3_dev_rx_descriptor_status(void *rx_queue, uint16_t offset);
+int hns3_dev_tx_descriptor_status(void *tx_queue, uint16_t offset);
+void hns3_tx_push_init(struct rte_eth_dev *dev);
+void hns3_stop_tx_datapath(struct rte_eth_dev *dev);
+void hns3_start_tx_datapath(struct rte_eth_dev *dev);
+
#endif /* _HNS3_RXTX_H_ */